You are a "nice" person, asking for help and blaming people who tried to help ... I get back my apologies and will not bother you anymore.Rainervs said:
P.S. I am using foobar for a long time, and used resampling in it many times ... right click on the file, convert->settings->necessary rate and bit depth, be sure that DSP mode with SSRC is active, then run conversion, single file output ...
I though that it is a good deal to help the person, working on religious projects
, but I was WRONG ... not everything that looks shiny, is gold ...Rainnervs, I am also thinking of making a software player which uses it's own sample-rate internal software conversion, including, direct to a 128x DSD stream (to bypass the DAC’s internal converter) for a true DSD usb2.0 interface hardware I'm designing.
If you don’t care about the CPU utilization % while playing just some .wav/direct CD playing, like 20-70% of a 1 Ghz PIII, I think a small fan-less 1GHz pc can serve as really good DSP hardware, or, just to play standard CDs on you PC, using a 64bit floating upsampling / downsampling codec. Let’s roast those 32/34 bit dedicated audio DSPs chipsets.
I do know how to program C and I can design software DSP codecs, my weakness is lack of Windows programming experience. Something I really don’t want to learn since I specialize in core functions/engines.
If you have further interest in the subject, email me privately.
If you don’t care about the CPU utilization % while playing just some .wav/direct CD playing, like 20-70% of a 1 Ghz PIII, I think a small fan-less 1GHz pc can serve as really good DSP hardware, or, just to play standard CDs on you PC, using a 64bit floating upsampling / downsampling codec. Let’s roast those 32/34 bit dedicated audio DSPs chipsets.
I do know how to program C and I can design software DSP codecs, my weakness is lack of Windows programming experience. Something I really don’t want to learn since I specialize in core functions/engines.
If you have further interest in the subject, email me privately.
Here is an oldy, if you only want to downsample your source material, try the following:
Get the old Cooledit 2000.
Just relying on in internal downsampling will only give you so-so results. Follow these steps & you should get significantly better performance.
-----------------------------------------
Load your audio.
Use the FFT audio filter, on the 96K, make a smooth rounded off low pass filter which is flat to 20Khz, the rolls off to 0% at 22Khz. Use 16Kpoint, or better on the FFT.
-SAVE-
Now, resample to 44Khz, 16 bit. Use the cheapest pre & post filtering.
-SAVE-
Redoo the same FFT audio filter on the 44Khz/16 bit sample. & save to .wav.
The FFT & resample will be VERY slow, like a few hours for a 10 minute track, but, you will get the quality.
-----------------------------------------
Get the old Cooledit 2000.
Just relying on in internal downsampling will only give you so-so results. Follow these steps & you should get significantly better performance.
-----------------------------------------
Load your audio.
Use the FFT audio filter, on the 96K, make a smooth rounded off low pass filter which is flat to 20Khz, the rolls off to 0% at 22Khz. Use 16Kpoint, or better on the FFT.
-SAVE-
Now, resample to 44Khz, 16 bit. Use the cheapest pre & post filtering.
-SAVE-
Redoo the same FFT audio filter on the 44Khz/16 bit sample. & save to .wav.
The FFT & resample will be VERY slow, like a few hours for a 10 minute track, but, you will get the quality.
-----------------------------------------
Rainervs said:
Argh, they will kill Cooledit Pro 🙁
I didnt notice Adobe had bought them out.
You can still get a shareware (limited features) free copy here:
http://www.adobe.com/products/tryadobe/main.jsp
MWP said:
The already killed it.....
Cooledit 96 note:
When using a 32bit os, NT4/Win2K/WinXP, the realtime spectrograph in CoolEdit 96 It the BEST for analysing audio & hidden artifacts, artifacts due to compression, or resampling, or that pesky 15.7Khz residual within recordings picked up from the yoke of you CRT monitors.
It blows away the CoolEdit 2000/CoolEdit pro, unless you use WinME/98/95, then it runs the same as the newer Cooledits. It's a bug, but, an extremely useful 1.
I have a friend who works in a studio as a sound engineer.
He uses Steinberg Wavelab. It includes the Apogee UV22HR dithering feature. This is, from what I understand, the highest quality dithering process generally available. They have some white papers that give a very detailed explanation. Even some audio clips to show the difference between difference dithering types. They are just coming out with v5. The v4 software can be had for roughly $300. I am sure if you buy now you can get a free upgrade to v5 when it become available.
He uses Steinberg Wavelab. It includes the Apogee UV22HR dithering feature. This is, from what I understand, the highest quality dithering process generally available. They have some white papers that give a very detailed explanation. Even some audio clips to show the difference between difference dithering types. They are just coming out with v5. The v4 software can be had for roughly $300. I am sure if you buy now you can get a free upgrade to v5 when it become available.
I still would put the 24/96 initially through a good FFT filter, before the downsampling, limiting the bandwidth to the CD range. It's not for altering the sound, but, to remove any previously unheard HF signals, usually ones caused by interfearance, not what is intended recording, which may leach their way through the downsampling algorythm & appear as tones, or whisles intertwined in the playback.
Brian Guralnick said:
Any downward resampling algorithm automatically does this filtering. No need to do it manually.
Brian Guralnick said:
Brian, You mean it's a feature not a bug. I never kept a copy of '96, could you clarify what you mean by the difference?
I had found that link on Adobe ... it gets you the full version of Audition 1.0, limited to 30 days after first launch. Sort of weird, considering they're now shipping version 1.5. Some of the features in 1.5 might actually make the product worthwhile (like the frequency domain editing).MWP said:
The resampling in Audition 1.0 is pretty good. It has idfferent artifacts than SSRC, and I'm not sure yet which one is better.
Anybody know where I can find the "real" cooledit?
Rainer
For all the fuss made over dithering, I cannot hear large differences caused by choice of dithering ... at least nowhere near as significant as the artifacts introduced by sample rate conversion.Cameron said:
Also, converting 96/24 to 44/24 introduces most of the problems; the trip from 44/24 to 44/16 is easy by comparison.
Apogee has certainly built the best reputation for its dithering algorithm. You might want to visit www.24-96.net/dither and try the blind comparison of 21 different dither methods. The one most preferred statistically was Extrabit (now renamed MBM). My personal preference was for the ones that used the least amount of audible noise -- they seemed to give the smoothest sound, surprisingly. Cooledit and SoundForge were 2 of my top 3 blind choices for dither.
All of which brings me no closer to finding a good resampling algorithm ...
Rainer
You're right, Gordon, I *am* a nice person most of the time ...Gordon McGregor said:
So, foobar2000 includes SSRC as its resampling engine. Didn't you notice the first reply in this thread, recommending SSRC? And the message after that, when I said I had tried SSRC, and that it seemed quite good?
And if this is your best recommendation, based on your years of personal expoerience, why wasn't it your first (or at least second) recommendation?
Rainer
Cameron said:
Yes, this should be true. I just prefer the manual Cooledit's FFT filter. Their internal pre/post filter when downsampling SUCKS, I'm not even sure why they don't use an FFT filter there.
scott wurcer said:
This feature is a bug. The old cooledit 96 running in a 32 bit enviroment runs the FFT window at 4x speed. An example when loading an .mp3, not only do CLEARLY see the first rolloff off in the upper frequencies, you even SEE all the DROPOUTS in the lower and mid frequencies demonstrating what most of the crappy encoders do. You also see a huge difference with the matrix stereo encoded .mp3 where many-many-many times, there are pure mono patches throughout mid & high frequencies throughout the audio.
With Cooledit96 & old 16 bit OSs, or, new cooledit 2000 & above, the analyser is way, way, way, way... too slow to easily make out any of these problems.
Note: I do still have the original demo version of CoolEdit 96 which I keep for the analysis purpose exclusively. It seems not to interfrear with the newer Cooledits...
I promised don't bother you anymore, and I will keep my promise.Rainervs said:
You're right, Gordon, I *am* a nice person most of the time ...
So, foobar2000 includes SSRC as its resampling engine. Didn't you notice the first reply in this thread, recommending SSRC? And the message after that, when I said I had tried SSRC, and that it seemed quite good?
And if this is your best recommendation, based on your years of personal expoerience, why wasn't it your first (or at least second) recommendation?
Rainer
Your question has an answer in the previous posts - I offered you the best solution for your particular situation, which is hardware, dCS972 or Lynx or something else ... http://www.stereophile.com/digitalsourcereviews/260/index8.html http://www.stereophile.com/digitalsourcereviews/260/index9.html(IMO), you declined it, as you wanted to use software; I offered the best solution for downconversion in software, which could be used with already discussed SSRC (for downsampling), for example, you declined it and wanted all-in-one for downsampling and downconversion, I offered foobar - all-in-one solution, and it supports kernel-stream processing (it is almost like hardware processing), you didn't even try it, as it was too complicated for you without the detailed manual 🙄 ... and started to blame me
... what else do you want? Already discussed CoolEdit 96, CoolEdit Pro or SoundForge, as Nuendo 2.0 or SADiE will be too expensive for you?
I don't see ANY reason for me to continue this discussion, sorry.
Just for the record ... SSRC is a COMPLETE solution to the problem of converting 96/24 to 44.1/16. It includes a Sample Rate Converter, and it includes a choice of dithering methods. There is no need to go looking elsewhere for other pieces. It will also do normalization, if you want it. All by itself, with no outside assistance. And, if you are so inclined, you can do just the sample rate conversion with SSRC, and use some other method for the dithering; but you don't have to.
The only thing wrong with SSRC (for me) is that it adds a certain "sharpness" to the final sound. But it does not make the sound muddy, like most of the other sample rate converters I have tried so far. When I finish comparing all the pertinent suggestions I have received here, I will post a comparison report.
Rainer
The only thing wrong with SSRC (for me) is that it adds a certain "sharpness" to the final sound. But it does not make the sound muddy, like most of the other sample rate converters I have tried so far. When I finish comparing all the pertinent suggestions I have received here, I will post a comparison report.
Rainer
SonicFoundry
SonicFoundry's Sound Forge has an EXCELLENT software resampler. (plus a full suite of powerful editing tools)
You can set several levels of precision, and even apply anti-aliasing to the resample. (way too cool!)
Dunno how bad Sony screwed up the company when they recently bought SonicFoundry, for sure the new site is a complete mess... but I can personally vouch for version 6.0 of Sound Forge as being truly excellent. (!!!) Right up your alley, and it's very affordable.
I have zero complaints with it and I've tried many other solutions in the past.
Vegas is powerful too, but I think for 2 channel audio, sound forge is the way to go.
Give SF a go. You won't be disappointed! (lots of ministries use it, and Sonic Foundry was very generous in sponsoring ministries... dunno about heathen sony though!)
And for a hardware solution, Lynx and RME are top notch products!
(However, RME has the edge I think, because they use "arrays" instead of dedicated DSP's, so they are likely to outlive DSP solutions, mostly because they are easily reprogramable, and way faster than DSP's. But Lynx is VERY good too! Both are light years ahead of mediocre M-Audio "solutions")
And foobar? Excellent, but a bit advanced. GREAT player when it comes to fidelity! I use it!
Best Regards,
Head-Spaz
<><
SonicFoundry's Sound Forge has an EXCELLENT software resampler. (plus a full suite of powerful editing tools)
You can set several levels of precision, and even apply anti-aliasing to the resample. (way too cool!)
Dunno how bad Sony screwed up the company when they recently bought SonicFoundry, for sure the new site is a complete mess... but I can personally vouch for version 6.0 of Sound Forge as being truly excellent. (!!!) Right up your alley, and it's very affordable.
I have zero complaints with it and I've tried many other solutions in the past.
Vegas is powerful too, but I think for 2 channel audio, sound forge is the way to go.
Give SF a go. You won't be disappointed! (lots of ministries use it, and Sonic Foundry was very generous in sponsoring ministries... dunno about heathen sony though!)
And for a hardware solution, Lynx and RME are top notch products!
(However, RME has the edge I think, because they use "arrays" instead of dedicated DSP's, so they are likely to outlive DSP solutions, mostly because they are easily reprogramable, and way faster than DSP's. But Lynx is VERY good too! Both are light years ahead of mediocre M-Audio "solutions")
And foobar? Excellent, but a bit advanced. GREAT player when it comes to fidelity! I use it!
Best Regards,
Head-Spaz
<><
Re: SonicFoundry
I respectfully disagree concerning superiority of RME vs Lynx Two B or L22. Here is the link with measurements http://audio.rightmark.org/test_results.shtml , and I know what chips used in both cards, therefore with the limitations of elements, RME, even being one of the top competitors to Lynx 2 series, is a bit worse, IMO.
Other serious competitors for Lynx 2 series are E-Mu 1212M/1820M with their low prices and almost the same ADC/DAC convertors and almost comparable op amps.
Concerning the speed, don't forget about the CPU loading as well ... DSP, used in Lynx, is very fast one - Xilinx Spartan 2, http://direct.xilinx.com/bvdocs/publications/ds001.pdf
fast enough even for Firewire 400mbs signals ... this DSP significantly lowers the loading of the main CPU in the PC.
Lynx 2 uses it's own in-system programmable PROM XC18V02 http://sunset.roma1.infn.it/muonl1/system/comp_datasheets/xc18v00.pdf, therefore is not "less programmable", then RME solutions. Also Lynx 2 has superior level of jitter in inner clock, peak-to-peak level of it, measured with Audio Precision system, is less then 20ps in the range of 400Hz-20kHz.
http://www.ixbt.com/multimedia/lynx-two/plata.jpg
My personal mesurements for Lynx L22 in 16/44 mode are even better (analog loopback):
Frequency response (from 40 Hz to 15 kHz), dB: +0.01, -0.03 Excellent
Noise level, dB (A): -96.6 Excellent
Dynamic range, dB (A): 96.2 Excellent
THD, %: 0.0021 Excellent
IMD, %: 0.0069 Excellent
Stereo crosstalk, dB: -97.7 Excellent
As we can see, the level of limitation for 16 bit signal (16 bit = 96.3dB) is reached even in analog mode (1% is the max measuring error in RMAA). RME in such mode is just a bit, but still worse.
Hi Spaz,head_spaz said:
I respectfully disagree concerning superiority of RME vs Lynx Two B or L22. Here is the link with measurements http://audio.rightmark.org/test_results.shtml , and I know what chips used in both cards, therefore with the limitations of elements, RME, even being one of the top competitors to Lynx 2 series, is a bit worse, IMO.
Other serious competitors for Lynx 2 series are E-Mu 1212M/1820M with their low prices and almost the same ADC/DAC convertors and almost comparable op amps.
Concerning the speed, don't forget about the CPU loading as well ... DSP, used in Lynx, is very fast one - Xilinx Spartan 2, http://direct.xilinx.com/bvdocs/publications/ds001.pdf
fast enough even for Firewire 400mbs signals ... this DSP significantly lowers the loading of the main CPU in the PC.
Lynx 2 uses it's own in-system programmable PROM XC18V02 http://sunset.roma1.infn.it/muonl1/system/comp_datasheets/xc18v00.pdf, therefore is not "less programmable", then RME solutions. Also Lynx 2 has superior level of jitter in inner clock, peak-to-peak level of it, measured with Audio Precision system, is less then 20ps in the range of 400Hz-20kHz.
http://www.ixbt.com/multimedia/lynx-two/plata.jpg
My personal mesurements for Lynx L22 in 16/44 mode are even better (analog loopback):
Frequency response (from 40 Hz to 15 kHz), dB: +0.01, -0.03 Excellent
Noise level, dB (A): -96.6 Excellent
Dynamic range, dB (A): 96.2 Excellent
THD, %: 0.0021 Excellent
IMD, %: 0.0069 Excellent
Stereo crosstalk, dB: -97.7 Excellent
As we can see, the level of limitation for 16 bit signal (16 bit = 96.3dB) is reached even in analog mode (1% is the max measuring error in RMAA). RME in such mode is just a bit, but still worse.
- Status
- Not open for further replies.